MobtgZhang

About Me

I am a Ph.D. student in Electronic Information at the School of Computer Science and Technology, Fudan University. My research interests lie in LLMs, with a particular focus on pre-training data curation, instruction data synthesis, agent trace data curation, knowledge management, and the cognitive capabilities of language models. Recently, I have been focusing on the following questions:

  • Intelligent Agent Workflow: How can intelligent agents plan, execute, and adapt to complex tasks across different domains and environments? I am particularly interested in task planning, analogical reasoning, structured memory, and compositional workflow synthesis. My research explores how agents can leverage past experiences, identify transferable task structures, and construct new workflows for unfamiliar tasks, rather than relying solely on direct trajectory retrieval.
  • Agent Data-Centric Curation: How can we curate high-quality agent interaction data to improve the learning and generalization capabilities of intelligent agents? I focus on the collection, selection, organization, and synthesis of agent trajectories and tool-use traces. In particular, I am interested in identifying valuable interaction patterns, preserving structural and contextual information, and transforming fragmented execution experiences into reusable training data.

Education

Fudan University

  • Ph.D. in Electronic Information (Sep. 2024 – Present)
  • School of Computing and Intelligence Innovation
  • Research interests: Large Language Models(LLMs), data curation, and LLM Cognition.

Shanghai University of Engineering Science

  • M.S. in Control Science and Engineering (Sep. 2021 – Jun. 2024)
  • School of Electronic and Electrical Engineering.
  • Research interests: Knowledge graph validation, knowledge management, and language models.

Shanxi University

  • B.S. in Computer Science and Technology (Oct. 2015 – Jul. 2019)
  • School of Computer and Information Technology.
  • Research interests: Natural Language Process.

Publications

  1. Geng Zhang, Yizhou Ying, Sihang Jiang, Jiaqing Liang, Guanglei Yue, Yifei Fu, Hailin Hu, and Yanghua Xiao. From Remembering to Metacognition: Do Existing Benchmarks Accurately Evaluate LLMs? Findings of ACL: EMNLP 2025, pp. 13440–13457, 2025.

  2. Geng Zhang, etc. al. Metacognitive Memory: Teaching Agents to Know What They Don’t Know. Findings of WSDM 2027.

  3. Yizhou Ying, Geng Zhang, Cui Danxin, Chengyu Du, Guanglei Yue, Sihang Jiang, Jiaqing Liang, Yifei Fu, Hailin Hu, and Yanghua Xiao. Data-Efficient Selection via Grammatical Complexity in Continual Pre-training of Domain-Specific LLMs. EMNLP 2025, pp. 22055–22069, 2025.

  4. G. Zhang, Y. J. Xiong, J. P. Hu, et al. Triplet Trustworthiness Validation with Knowledge Graph Reasoning. Engineering Applications of Artificial Intelligence, 141, 109813, 2025.

  5. Geng Zhang and Jianpeng Hu. Enhanced Industrial Text Classification via Hyper Variational Graph-Guided Global Context Integration. PeerJ Computer Science, 10, e1788, 2024.

  6. Guo X., Zhang G., Wang S., et al. Multi-Way Matching Based Fine-Grained Sentiment Analysis for User Reviews. Neural Computing and Applications, 32(10), 5409–5423, 2020.

Interests

I enjoy exploring a wide range of topics in computer science, with a particular passion for operating systems, Large Language Models (LLMs), and FPGA research. Beyond computer science, I have a strong interest in mathematics, especially modular forms, functional analysis, and mathematical analysis. I enjoy exploring mathematical theories and appreciating the beauty of abstract mathematical structures. I am also fascinated by history. I particularly enjoy analyzing historical events, understanding their underlying contexts, and examining their long-term impacts on society and the world.